Rainfall is an important factor for agricultural cultivation. Therefore, astudy of past rainfall records using statistical techniques to forecast the rainfallwill be essential for planning and setting policy in agricultural extension. This isone responsibility of the Center of Agricultural Economics, Ministry of Agricultureand Co-operatives. The aims of this project are as follows: I. To compare four methods of forecasting, namely: Constant Mean Model,Decomposition Method, Winters's Forecast Method and Box-Jenkins Technique. Data usedto forecast the rainfall in 1997 are taken from the past rainfall records of1986-1996 in the 19 provinces of Northeast Thailand: Loei, Udon Thani, Nong Khai,Nong Bua Lamphu, Sakon Nakhon, Nakhon Phanom, Chaiyaphum, Khon Kaen, Kalasin,Mukdahan, Mahasarakam, Roi Et, Yasothon, Amnat Charoen, Ubon Ratchathani, Sisaket,Surin, Buriram and Nakhon Ratchasima. Data which gives the lowest Mean AbsolutePercent Error (MAPE) will represent the most suitable method. Results show that theDecomposition Method is the most suitable method for all provinces in NortheastThailand. II. Trend Surface Analysis was conducted using 1996 rainfall records and thepositions (Latitude, Longitude) of 93 selected stations (out of a total of 300) in l9provinces to predict the rainfall in 10 targeted stations. The forecasting valuesobtained from these 10 stations compare Trend Surface Analysis with the Constant MeanModel, Decomposition Method, Winters's Forecast Method and Box-Jenkins Technique.They show that the Trend Surface Analysis provides the most suitable forecasting dataof all the methods.